Let's visualize GraphConvModel of DeepChem

Introduction

I wanted to start Deep Learning with a compound, so I hacked the GraphConvModel of DeepChem last time and output it as text by the summary method.

This time I tried to visualize it with plot_model.

environment

• DeepChem 2.3
• pydot

Method

If you put the following code in line 624 of the file where the GraphConvModel class definition is made and create a prediction model with appropriate data, a file called model.png will be generated.

<anaconda3>/envs/deepchem/lib/python3.7/site-packages/deepchem/models/graph_conv.py

    from tensorflow.keras.utils import plot_model
    plot_model(
            model,
            show_shapes=True,
        )

result

Like this. The order of Input is changed and it is hard to see, but it may be easier to imagine than the text-based one. Weave is going to be more complicated.